A new adaptation algorithm for active control of sound

In this paper, a new adaptation algorithm which requires no knowledge of the error signal samples is presented. It is specially useful in applications such as active noise control where a filtered version of the error signal rather than the error signal itself is available. This algorithm is based on a technique for multivariate function minimization. It only needs as feedback signal an estimate of the error signal power. The fact that the error signal has been filtered by an unknown filter does not affect the stability properties of the algorithm. Due to that fact, this algorithm becomes a better choice than the classical filtered-X LMS based versions. A comparison between this algorithm and LMS is also presented. Simulations and real time experiences validate the authors' conclusions.